RSNA 2005

Outline
Parallel Imaging: Physical principles Implementation: SENSE vs SMASH New technical advances: Multicoilmultichannel hardware Parallel Imaging: Practical uses

Parallel Imaging Made Easy

Bachir Taouli, MD
Assistant Professor

Vivian S. Lee, MD, PhD

Professor and Vice-Chair of Research Department of Radiology New York University Medical Center

www.med.nyu.edu/mri
Nov 2005

Parallel Imaging Nomenclature

SENSE = SENSitivity Encoding ASSET = Array Spatial & Sensitivity Encoding Technique SPEEDER SMASH = SiMultaneous Acquisition with Spatial Harmonics GRAPPA = GeneRalized Autocalibrating Partially Parallel Algorithm iPAT = integrated Parallel Acquisition Technology

MR Physics: A Review

Physics Review

Physics Review

= Number of phase-encoding steps

Spatial localization is achieved by:

Repeated acquisitions of echoes with different frequency-encoding and phase-encoding gradients Acquisition Time = TR x Number of Phase-Encoding Steps
From Lee VS, Cardiovascular MRI: Physical Principles to Practical Protocols, Lippincott 2005

Physics Review: k-space

Physics Review: k-space

K-space

MRI

Periphery Periphery

K-space is the raw data space Fourier transform of k-space gives the image Center of k-space = image contrast Periphery of k-space = fine details

Physics Review: k-space

The spacing of data in k-space is inversely related to the size of the image

Center Center

Physics Review: recFOV

Rectangular FOV = Undersampling of phaseencoding steps Wider spacing in k-space = narrower FOV Fewer phase-encoding steps = faster acquisition

Parallel Imaging

Back to Parallel Imaging

Implementation: Undersample phase-encoding lines in k-space and use coil information to recover data

Parallel Imaging
Concept: The strength of a signal detected by one coil versus another can also be used for spatial localization

Parallel Imaging
Two methods
Unwrap the image (SENSE, ASSET, SPEEDER)

Parallel Imaging
Undersampling of phase-encoding steps (NPE) directly translates into shorter acquisition times
Acquisition time = TR x NPE

Parallel Imaging
Take Home Message #1:
Parallel imaging is like rectangular FOV to the extreme, without wraparound artifact

Undersampling factor = R (reduction factor) E.g. Undersample by 50% R = 2

Coils and Receivers

Phased array coils have multiple coil elements
Body phased array coils: typically 4 6 for anterior and 4 6 for posterior

Parallel Imaging
Maximum R depends on
Number of coils (with separate receivers) in the PHASE-ENCODING direction

MR systems have separate receiver channels

Typically 4 8 receiver channels on most systems Recently 32 channel systems have come to market

multiplex

Multiple One MultipleCoils Coils Onechannel channel One CoilOne channel

PARALLEL PARALLEL IMAGING! IMAGING!

Parallel Imaging
Body imaging with two 3 x 2 arrays (12 channels)
Axial imaging (PE = A - P): R = 2 Sagittal imaging (PE = A - P): R = 2 Coronal imaging (PE = L - R): R = 3

Parallel Imaging
R reduces acquisition times for 2D and 3D Acquisition time reduced by a factor of about R
Undersample NPE 50% R = 2 Undersample NPE 33% R = 3 Undersample NPE 25% R = 4

For 3D imaging, with NPE1 and NPE2,

Undersample NPE1 50%, NPE2 50% R=4 Undersample NPE1 33%, NPE2 50% R=6 Undersample NPE1 33%, NPE2 33% R=9

Parallel Imaging

Take Home Message #2:

To do Parallel imaging, you need:
Coils with multiple elements Matching multichannel receiver systems

Coil Sensitivity Profiles

Separate acquisition for coil sensitivity mapping problematic for breath-hold applications Integrated coil sensitivity mapping much better

Implementation Issues

Coil sensitivity profiles Limited rectangular FOV New coil designs Multichannel systems

Coil Sensitivity Profiles

Integrated approach
Full k-space sampling for central lines (e.g. 16-24 lines fully sampled) Central lines used for coil sensitivity map Theoretical reduction in acquisition time not achievable
256 x 128 matrix R = 2, should mean NPE = 64 (1/2 acq time) With central 24 lines for coil sensitivity map, NPE = 76 (acquisition time = 0.6 x original)

Coil Sensitivity Profiles

Parallel Imaging

Practical Tip:
Use manufacturer-provided pads or cushions to distance coils from the patients skin Reduces super-high sensitivity at skin Improves coil sensitivity profiles

Take Home Message #3:

Coil sensitivity profiles are needed for all parallel imaging and reduce its theoretical efficiency

Parallel Imaging and recFOV

Can be performed in conjunction with recFOV Tight recFOV causes wrapping in the image

Parallel Imaging and recFOV

Can be performed in conjunction with recFOV Tight recFOV causes wrapping in the image

Technical Challenges
Requires special coil designs with large number of independent elements and receiver channels for each Need coil sensitivity maps Computationally intensive Reduced SNR (R)

New Coil Designs

More coil elements to increase R Balance against reduced depth sensitivity Each element with its own preamplifier

TODAY: TODAY:

Philips 4 element (ant + post = 8)

GE 4 element (ant + post = 8)

Siemens 6 element (ant + post = 12)

New Coil Designs

FUTURE: FUTURE:

New MR system designs

All manufacturers have 8 channel systems, moving toward 16 & 32 36 Channels 76 Coil Elements

16 16 Channel Channel Coupled Coupled Array Array

64 64 Element Element Array Array Single Single echo echo imaging imaging Steve Wright et al Proc,2nd Joint EMBS/BMES, 2002

Ray RayLee, Lee,NYU NYU

(Siemens Avanto)

Outline
Parallel Imaging: Physical principles Implementation: SENSE vs SMASH New technical advances: Multicoilmultichannel hardware Parallel Imaging: Practical uses

Parallel Imaging: In Practice

Advantages of parallel imaging:
Decreased acquisition times Increased spatial resolution Increased temporal resolution Improved single shot imaging Improved contrast in IR sequences

In practice, these translate into

Van der Brink JS, et al. Eur J Rad 2003; 46:3 - 27

Parallel Imaging: In Practice

Improved routine body protocols
Faster (shorter breath-holds) More slices per acquisition Improved spatial resolution

Parallel Imaging: In Practice

Improved routine body protocols
Faster (shorter breath-holds) More slices per acquisition Improved spatial resolution

Some sequences newly practical in BH Functional body MR

Liver perfusion Renal function

Some sequences newly practical in BH Functional body MR

Liver perfusion Renal function

Liver/MRCP Protocol
No NoParallel Parallel Imaging Imaging
Scout STIR x 2 T1 GRE (dual echo) Coronal HASTE Axial HASTE Thick Slab 2D TSE Gd-3D GRE (VIBE) All 256 matrix 20 sec 19 sec x 2 20 sec 20 sec 20 sec 8 sec x 3 20 sec

Fat-Suppressed T2-weighted MRI

For breath-hold acquisitions
STIR (10 slices/19 sec breath-hold)

With WithParallel ParallelImaging Imaging

Scout 20 sec FS-FSE 13 sec T1 GRE (dual echo) 12 sec Coronal HASTE 20 sec Axial HASTE 20 sec 3D TSE 2 min Gd-3D GRE (VIBE) 14 sec Most 512 matrix (VIBE 320) R=23

STIR

FS-FSE

FS-FSE (20 slices/13 sec), R = 2

Breath-hold dual echo T1 GRE

VIBE with Parallel Imaging

VIBE (interpolated FS 3D GRE)
320 matrix, 96 partitions x 2 mm; 1.2 x 1 x 2 mm R=2 14 sec acquisition

In-phase and out-of-phase T1 GRE 12 sec acquisition for 20 slice positions 512 matrix R=2

Axial source images

Coronal reformatted images

New Liver/MRCP Protocol

Scout 20 sec FS-TSE T2 14 sec T1 GRE (in/out phase) 12 sec Coronal HASTE (MRCP) 20 sec Axial HASTE (MRCP) 20 sec 3D TSE (MRCP) 2 min Gd-3D GRE (VIBE) 14 sec Less respiratory artifact, better patient tolerance Improved image quality

Improved MRCP

Conventional 2D Thick Slab Dx: Ductal IPMT

3D TSE (< 2 min) With respiratory gating R=3 1 mm slice x 56

Improved MRCP
3D Fast Spin Echo
Single-shot or Two-shot Respiratory gating (navigator) One partition per respiratory cycle Total acquisition time = 2 5 min

Respiratory Triggering: Navigator

For non-breath hold acquisition
Navigator integrated into acquisition Within narrow acceptance window data acquired Higher resolution imaging or noncompliant patients

Respiratory-Triggered 3D MRCP

Breath-hold 3D MRCP!
3D TSE with Parallel imaging!

3D 384 matrix 60 x 1 mm slices 2 min

3D PACE TSE (2 min)

Zhang J et al, ISMRM 2004

3D BH TSE (30 sec)

Pelvic MRI
Protocol
T2-weighted FSE, two planes T1-weighted gradient echo (dual echo) 3D GRE (VIBE) Gd

Pelvic MRI with Parallel Imaging

512 Matrix 4.0 mm

R=2
512 Matrix 4.0 mm

6:16

3:08

Pelvic MRI with Parallel Imaging

Parallel Imaging: In Practice

Routine body protocols
Faster (shorter breath-holds) More slices per acquisition Improved spatial resolution

Some sequences newly practical in BH Functional body MR

512 Matrix 4.0 mm x 35 TA 2:00 512 Matrix 4.0 mm x 35 TA 2:36

Liver perfusion Renal function

R=2

Liver Diffusion-Weighted Imaging

Background
Tests to diagnose early fibrosis limitedbiopsy usually required Decreased hepatic apparent diffusion coefficient (ADC) in liver cirrhosis Explanation: restricted water diffusion related to liver fibrosis?

Liver Diffusion-Weighted Imaging

Parallel Imaging and liver DWI
Reduced number of phase-encoding steps Shorter TE (shorter echo train) with single shot sequence >>> better SNR in the liver Improved image quality-less distortion

Technical challenges
Long acquisition times (esp DTI) Long TEs with susceptibility artifacts

DTI with Parallel Imaging

Conventional Diff. SENSE

Parallel Imaging: In Practice

Routine body protocols
Faster (shorter breath-holds) More slices per acquisition Improved spatial resolution

DTI-SENSE

b=0

b = 500

ADC

SENSE-CDI Normal (n = 10) HCV (n = 5)** P

1.52 0.15 (1.28-1.80) 1.17 0.22 (0.721.39) <0.006

SENSE-DTI
1.51 0.21 (1.271.99) 1.24 0.20 (1.171.35) <0.03

Makes sequences newly practical in BH Functional body MR

Liver perfusion Renal function

Taouli et al, ARRS 2004

Whole Organ Perfusion

Fast dynamic contrast-enhanced 3D with parallel imaging Applications:
Organ function/physiology
Liver perfusion for cirrhosis Kidney function
Development of Metastases

Liver Malignancy
Tumor cell

Development of HCC
Dysplastic nodule Small HCC HCC

Endothelial cell Tumor cell

Cancer detection and characterization

Breast imaging Liver perfusion for metastases and hepatocellular carcinoma

Extravasation of circulating tumor cell

Micrometastasis

Metastasis

Progressive Increase in Arterial Supply

Unpaired Arteries
Neoangiogenesis in dysplastic nodules and HCC
Normal and cirrhosis: 0 unpaired arteries Low grade DN < high grade DN < HCC (p<.00001)

Whole Liver Perfusion

Tumors and cirrhosis have preferential arterial enhancement 3D perfusion may improved diagnosis

Park et al. Am J Surg Path 1998; 22:656 Krinsky et al Radiology 1998; 209:461

3.6 sec VIBE Dynamic Liver Perfusion

Whole Liver Perfusion

4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0 10 20 30 40 50 60

Renal MRI
Commonly used for anatomic studies
Renal masses Renal artery stenosis Renal transplants

[Gd]

Aorta_cor PV_cor Liver_cor

time (sec)

Potential functional applications

Single kidney GFR Renovascular disease Renal transplant dysfunction

dCL/dt = k1aCa(t-tdelay) + k1pCp(t) - k2CL(t) k1a k1p k2

Liver

Perfusion Map

Ultra-low Dose MR Renography

Ultra-low doses of Gd-DTPA (0.005-0.01 mmol/kg) result in visible corticomedullary patterns of enhancement

Functional Studies
Gadolinium contrast filtered at glomerulus without tubular secretion or resorption (like inulin or creatinine) No nephrotoxicity Low dose Gd kinetics through kidney
Aorta kidney (RPF) Cortex medulla (GFR) Medulla pelvis (tubular function)

Lee et al. Radiology 2001; 221:375

Results: Normal Cr
Medullary enhancement pattern: sharp early rise within 20 sec (vascular phase) Second rise within 2 min (GFR)
1.4 1.2 1 C ortex M edulla

Results: Abnormal Cr
1 .4 1 .2 1 0.8 0.6 0.4 0.2 0 0 22 44 66 88
Cor t ex Medulla

Relative SI

Relative SI

0.8 0.6 0.4 0.2 0 0 22 44 66 88 110 132 154 176 198 220

Tim e (sec)

1 1 0

1 32 1 54 1 76 1 98 220

Elevated serum creatinine: Absence of second medullary peak

Time (s )

Left RAS
Baseline

Left RAS
Baseline Left Kidney (CP)
60 50 Signal Intensity 40 30 20 10 0 4 20 36 52 68 84 100 116 132 148 164 180 196 212 Tim e (sec) 228

1.58

Post ACE-I Left Kidney (CP)

1.15

Signal Intensity

ACE-I

70 60 50 40 30 20 10 0
2 92 11 0 12 8 14 6 16 4 18 2 20 0 21 8 23 6 56 20 38 74

Tim e (sec)

Left RAS
Baseline Left Kidney (CP)
60 50 Signal Intensity 40 30 20 10 0 4 20 36 52 68 84 100 116 132 148 164 180 196 212 Tim e (sec) 228

Kidney Function: 3D
1.58

Use Gd-DTPA (2-4 ml) as tracer for kidney function Fast high-resolution 3D imaging with parallel imaging

Post ACE-I Left Kidney (CP)

70 60 Signal Intensity 50 40 30 20 10 0

1.15

Tim e (sec)

3 sec acquisition - 2 ml Gd 48 x 2 mm (R = 3 Acquisition 6/18

Kidney Function: 3D
Use Gd-DTPA (2-4 ml) as tracer for kidney function Fast high-resolution 3D imaging with parallel imaging

92 11 0 12 8 14 6 16 4 18 2 20 0 21 8 23 6

20

38

56

74

Summary
Parallel Imaging in Body MRI means
Faster, better routine clinical imaging Enhanced functional, physiologic imaging New applications previously not possible

3 sec acquisition - 2 ml Gd 48 x 2 mm (R = 3) Acquisition 6/18

Selected slice from 3D 0 - 4 minutes 18 3D Acquisitions

www.med.nyu.edu/mri

International Society for Magnetic Resonance in Medicine (ISMRM)

Seattle, Washington (USA) May 6-12, 2006 www.ismrm.org
2005 2005

RSNA 2005

Parallel Imaging Made Easy

Bachir Taouli Assistant Professor of Radiology
Vivian S. Lee, M.D., Ph.D. Professor and Vice-Chair of Research Department of Radiology New York University Medical Center

www.med.nyu.edu/mri
Nov 2005